Common knowledge was anyways acquired through the open internet. So how does it matter how many layers of distillation happens. Proprietary data and knowledge will see a local version of the model keeping that will belong to the very companies that own this data. Also the simulation data doesn’t work in the last mile for the enterprise. It was hyped up data generators to begin with.
#EnterpriseAI
RIDICULOUS - The White House is discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models
There is an easy way to get around this
Just distill a Chinese model and create a US open-weights version.
This will instantly become a big business. Open-weights can't be stopped easily
@clamanna Thanks for the details @clamanna! Looks like the issue repeated consistently across all servers with this load footprint and i trust you that the load testing framework will also see changes after this event.
Very valid points made here. Also, its important to understand how critical production software systems are in a world that thinks software engineering is dead.
#MicrosofOutlookDown#Microsoft365
Half the planet just ground to a complete halt because Microsoft pushed a broken routine update to Entra ID. Outlook dead, Teams refusing connections, SharePoint and OneDrive throwing sync errors everywhere. Over 300 million users across more than 100 countries were essentially locked out of doing their jobs.
The chaos kicked off around 9:30 AM Pacific Time right during peak enterprise business hours across the Americas and Europe. What followed was nearly eight straight hours of absolute disruption across global corporate communication channels. Millions of emails bounced with generic service unavailable errors while critical video meetings dropped mid-call.
Once again, this was not some sophisticated nation-state cyberattack or a catastrophic external breach. It came down to an internal engineering deployment that misconfigured core access policies and authentication broker services. When identity verification fails at the Entra ID layer, every single dependent productivity app collapses behind it like dominoes.
It took Microsoft engineering teams until roughly 10 PM PT to progressively stabilize systems and bring services back online, leaving thousands of companies paralyzed for an entire work day. Nobody could coordinate, nobody could access critical files, and support queues were utterly swamped with automated responses and useless generic status notices.
The sheer fragility of centralizing every critical enterprise workflow under a single proprietary identity stack is staggering. When one cloud provider controls your email, video calls, calendars, and secure document storage, an internal misconfiguration on their side immediately becomes a total global operational shutdown for everyone else. Multi-cloud redundancy cannot remain just an abstract corporate talking point anymore.
Great to hear your thinking here. Many vcs in the valley still think that workflows are for fdes or SIs only and no business user/non-tech person will have the time to spend on building their own workflow. But I have first hand seen this feeling of being powerful and i know the business domain best and dont need to depend on others twice now in my career. Once at Microsoft with bringing low-code apps and now with Brackett enabling business agents for business users.
Unbelievable results coming out of Texas! And its hilarious to see how some wannabe ai influencers talking about midterms and their AI agenda washing on X. Focus on people and listen to what the American people have said through their ballots. #Midterms2026
This is a great time to build in the application stack! Also what’s amazing is that most of the large labs are now finally introducing concepts of “teach”. “Instruct” was the old paradigm. “Teach” will be the new one and when it works in real time that would be amazing! #whyweexist #llms #teachmachines #snt
3 frontier models in one day!
- Grok 4.6: Fable 5-level, but 85% cheaper.
- Qwen3.8-Max: 2.4T params, 95B active. Weights are out.
- DeepSeek-V4-Pro-0813: weights could drop any time. Heard it’s good, not just benchmaxxing.
Competition is great for consumers and businesses.
Releasing the model weights and technical report of Kimi K3.
Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.
New model architecture: 2.5x the intelligence per unit of compute, not just more params.
Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.
Model weights: https://t.co/7m7eEg6Y0B
Tech report: https://t.co/yeu6cjpMCT
Tech blog: https://t.co/YTfiMSNM1f
The only robbing that’s been happening post the #argentina#egypt match is the not acknowledging the grit of #argentina as a team and the genius of the GOAT #messi. Egypt played a great game but could not capitalize on their lead and Messi showed the world why he is the GOAT.